Paving ship self-adaptive displacement control system and method based on flow feedforward

Abstract

The invention relates to the technical field of ship displacement control, and particularly discloses a paving ship self-adaptive displacement control system and method based on flow feedforward, and the system comprises a cooperative data collection module which is used for responding to a displacement instruction of a paving ship, collecting local data of the paving ship and cross-equipment cooperative data, and sending the local data to a controller; carrying out delay compensation on the material flow data in the cross-equipment collaborative data; and the dynamic disturbance model construction module is used for constructing a dynamic disturbance model according to the local data of the paving ship and the cross-equipment collaborative data after delay compensation, and calculating and generating feed-forward compensation amounts for a side thruster and an anchor winch through the dynamic disturbance model. According to the method, accurate delay pre-compensation is carried out on cross-equipment collaborative material flow data, the dynamic disturbance model is constructed in combination with various internal and external environment variables, the system can pre-judge comprehensive disturbance in advance and generate the feed-forward compensation amount, and the hysteresis problem of traditional feedback control is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of ship displacement control technology, and in particular to an adaptive displacement control system and method for pavers based on flow feedforward. Background Technology

[0002] With the continuous development of underwater engineering technology, paving vessels play a crucial role in high-precision underwater operations such as laying crushed stone foundations for cross-sea immersed tunnels. During operation, paving vessels require precise displacement control through actuators such as side thrusters and anchor winches to ensure uniform material spreading. Due to the extremely harsh and complex marine operating environment, the vessel is not only subject to strong external environmental disturbances such as ocean currents and waves, but also to the intertwined effects of internal operating condition disturbances such as sudden changes in material pump flow and pipeline pressure fluctuations. Traditional displacement control schemes mostly rely on single hysteresis feedback regulation or simple feedforward compensation based on ideal conditions, which are difficult to adapt to complex operating conditions with superimposed multi-source dynamic disturbances.

[0003] In practical operation, existing technologies face prominent problems of strong variable coupling and overcompensation conflicts:

[0004] On the one hand, the material flow data transmitted across devices has inherent physical transmission and communication delays, which contradicts the system's real-time response requirements to transient disturbances, causing feedforward instructions to often be out of sync.

[0005] On the other hand, under extreme conditions, such as sudden strong winds and waves, the violent impact of external waves can directly cause abrupt changes in the ship's draft and nozzle attitude, resulting in severe pressure pulses in the internal material pipelines. Existing systems typically treat these abrupt changes in multiple variables caused by a single physical factor as independent disturbances and linearly superimpose them, leading the system to misjudge that it is subjected to a huge disturbance torque superimposed by multiple factors.

[0006] This overcompensation conflict caused by strong coupling between internal and external variables can lead to the system issuing extreme thrust commands, which not only causes the hull to vibrate violently in the opposite direction and deviate from the predetermined operating trajectory, but also easily causes overload and damage to the core actuators. Summary of the Invention

[0007] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose an adaptive displacement control system and method for pavers based on flow feedforward, to improve the displacement accuracy of pavers and the stability of construction operations under complex and harsh sea conditions.

[0008] To achieve the above objectives, a first aspect of the present invention proposes an adaptive displacement control system for pavers based on flow feedforward, comprising:

[0009] The collaborative data acquisition module is used to collect local data of the paving vessel and cross-device collaborative data in response to the paving vessel's movement command, and to perform delay compensation on the material flow data in the cross-device collaborative data.

[0010] The dynamic disturbance model construction module is used to construct a dynamic disturbance model based on the local data of the paving vessel and the cross-equipment collaborative data after delay compensation, and to calculate and generate feedforward compensation amounts for the thrusters and anchor winches through the dynamic disturbance model.

[0011] A two-stage adaptive decision-making module is used to generate control commands based on the feedforward compensation amount and feedback decoupling optimization logic to adjust the actual output of the side thruster and the anchor winch.

[0012] The calculation process for delay compensation of material flow data in the collaborative data acquisition module includes: obtaining the preset material pipeline length, real-time material flow rate and sensor inherent delay, and calculating the total measurement delay; calculating the predicted material flow rate at the actual moment of the disturbance based on the current material flow rate change rate and the total measurement delay, and inputting the predicted material flow rate into the dynamic disturbance model.

[0013] To achieve the above objectives, a second aspect of the present invention proposes an adaptive shift control method for pavers based on flow feedforward, comprising the following steps:

[0014] In response to the paving vessel's relocation command, local data of the paving vessel and cross-equipment collaborative data are collected, and delay compensation is performed on the material flow data in the cross-equipment collaborative data.

[0015] Based on the local data of the paving vessel and the cross-equipment collaborative data after delay compensation, a dynamic disturbance model is constructed, and feedforward compensation amounts for the side thrusters and anchor winches are calculated and generated through the dynamic disturbance model.

[0016] Based on the feedforward compensation and feedback decoupling optimization logic, control commands are generated to adjust the actual output of the side thruster and the anchor winch, and to execute the adaptive displacement of the paving vessel.

[0017] The calculation process for delay compensation of material flow data includes: obtaining the preset material pipeline length, real-time material flow rate and sensor inherent delay, and calculating the total measurement delay; calculating the predicted material flow at the actual moment of the disturbance based on the current material flow rate change rate and the total measurement delay, and inputting the predicted material flow into the dynamic disturbance model.

[0018] To achieve the above objectives, a third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described adaptive shift control method for pavers based on flow feedforward.

[0019] The adaptive displacement control system and method for pavers based on flow feedforward of this invention achieves precise pre-delay compensation for material flow data coordinated across equipment and constructs a dynamic disturbance model by combining various internal and external environmental variables. The system can predict comprehensive disturbances in advance and generate feedforward compensation, effectively overcoming the lag problem of traditional feedback control. Addressing control conflicts caused by strong coupling of multiple variables under extreme conditions, this solution can adaptively calculate the attenuation ratio and dynamically reduce the weight of characteristic weight coefficients by real-time monitoring of abrupt changes and sequence correlations of core external forces and internal states. This cleverly and accurately eliminates the superposition of false disturbances caused by strong physical coupling, fundamentally suppressing overcompensation and severe oscillations caused by torque misjudgment, and significantly improving the displacement accuracy, robustness, and stability of the entire construction chain of pavers under complex and harsh sea conditions. Attached Figure Description

[0020] Figure 1 This is a schematic diagram illustrating the implementation of the adaptive shift control system for pavers based on flow feedforward provided by the present invention.

[0021] Figure 2 This is a comparison curve of the effect of advance compensation for material flow data delay across equipment in the adaptive shift control system for pavers based on flow feedforward provided by the present invention.

[0022] Figure 3 This is a diagram showing the effect of material flow pulsation hierarchical filtering based on pressure derivative adaptive adjustment in the adaptive displacement control system for pavers based on flow feedforward provided by this invention.

[0023] Figure 4 This invention provides a K-means clustering spatial distribution and classification boundary partitioning diagram of multi-source disturbance characteristics under complex working conditions in the adaptive displacement control system for pavers based on flow feedforward.

[0024] Figure 5 This is a curve showing the dynamic game adjustment of feedforward and feedback weights in the energy consumption sensing-based two-order adaptive control system of the paving vessel adaptive displacement control system based on flow feedforward provided by the present invention.

[0025] Figure 6 This is a comparison diagram of the side thruster control command output before and after decoupling the strongly coupled wave and pressure state under extreme sea conditions in the adaptive displacement control system for pavers based on flow feedforward provided by the present invention.

[0026] Figure 7 This is a comparison chart of the smoothness of soft and hard switching of the feature weight parameters of the disturbance type transition period model in the adaptive displacement control system for pavers based on flow feedforward provided by this invention.

[0027] Figure 8 This is a flowchart illustrating the adaptive shift control method for pavers based on flow feedforward provided by the present invention.

[0028] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0029] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0030] The following description, with reference to the accompanying drawings, describes an embodiment of the adaptive shift control method, system, and electronic equipment for pavers based on flow feedforward.

[0031] Example 1:

[0032] In modern cross-sea immersed tunnel crushed stone foundation laying and large-scale underwater reclamation projects, paving vessels need to rely on complex propulsion and mooring equipment to achieve high-precision trajectory tracking and fixed-point displacement in harsh marine environments. The control system disclosed in this embodiment aims to completely solve the problems of position drift and overcompensation caused by multi-source disturbances, especially the strong coupling between material transport flow fluctuations and external sea conditions.

[0033] Specifically, the flow-feedforward-based adaptive displacement control system for pavers in this embodiment is applied to the central control unit of the paver. This central control unit is the brain of the entire vessel and typically consists of a high-performance industrial computer, a programmable logic controller, and a high-real-time fieldbus network.

[0034] like Figure 1 As shown, the system is physically and logically divided into several highly collaborative modules, mainly including a collaborative data acquisition module, a traffic interference processing module, a dynamic disturbance model construction module, a two-stage adaptive decision-making module, and a multi-dimensional anomaly processing module.

[0035] To achieve deep integration of feedforward and feedback control, the system first needs to acquire comprehensive and accurate operating condition data. Therefore, the system's collaborative data acquisition module responds to the paving vessel's movement commands, collecting local data from the paving vessel and cross-equipment collaborative data, and performing delay compensation on the material flow data within the cross-equipment collaborative data. In actual construction operations, the paving vessel often does not operate in isolation but needs to coordinate with the supply vessel, pressurization pump station, and shore-based monitoring center. The paving vessel's local data mainly includes real-time coordinates provided by the onboard global navigation satellite system, hull attitude provided by the inertial measurement unit, ship draft detected by the draft sensor, and real-time operational feedback status of actuators such as the thrusters and anchor winches. The cross-equipment collaborative data mainly includes the pump outlet pressure when the upstream supply system pumps materials, the material concentration at each node of the pipeline, and, most importantly, the material flow data.

[0036] Optionally, due to the long physical pipeline connecting the material supply pump station to the slab dumping nozzle on the paving vessel, and the time required for sensor data acquisition and bus packet transmission, the material flow data received by the central control unit exhibits a significant time lag. Directly using this delayed flow data for feedforward control would cause a time misalignment between the thrust generated by the propeller and the actual fluid reaction force reaching the dumping nozzle, failing to offset the disturbance and potentially exacerbating the vessel's sway. Therefore, the calculation process for delay compensation of material flow data in the collaborative data acquisition module comprises two main levels.

[0037] Specifically, the first level involves acquiring the preset material pipeline length, real-time material flow rate, and sensor inherent delay, and calculating the total measurement delay. The material pipeline length refers to the equivalent physical pipe distance from the upstream flow meter installation location to the final discharge port of the paving vessel; this length is typically a fixed, known parameter after deployment in a single operation. The real-time material flow rate is the physical movement speed of the mud-water mixture or aggregate slurry within the pipeline. The sensor inherent delay encompasses the electromagnetic flow meter's excitation response time, analog-to-digital conversion time, and the packet parsing time of the industrial Ethernet protocol.

[0038] For example, the process of calculating the total measurement delay in the collaborative data acquisition module includes: dividing the length of the material pipeline by the real-time material flow rate to obtain the fluid transmission delay, and adding the fluid transmission delay to the inherent delay of the sensor to obtain the total measurement delay.

[0039] The total measurement delay is calculated using the following formula:

[0040] ;

[0041] In the formula, Indicates the total measurement delay; Indicates the preset length of the material pipeline; Indicates real-time material flow rate; This indicates the inherent delay of the sensor.

[0042] By using physical modeling, the system accurately mathematically unifies the macroscopic fluid dynamic delay with the microscopic electronic communication delay, thereby deriving a time constant that can characterize the overall lag characteristics of the current system.

[0043] The second level involves calculating the predicted material flow rate at the actual moment of the disturbance's effect, based on the current material flow rate change rate and the total measurement delay, and then inputting this predicted material flow rate into the dynamic disturbance model. Since fluid transport is a physical process with continuous inertia, its short-term trend can be linearly extrapolated using the first derivative.

[0044] For example, the process of calculating the predicted material flow rate at the actual moment of the disturbance includes: multiplying the current material flow rate change rate by the total measurement delay to obtain the flow change compensation amount, and adding the real-time material flow rate at the current moment to the flow change compensation amount to obtain the predicted material flow rate.

[0045] The predicted material flow rate is calculated using the following formula:

[0046] ;

[0047] In the formula, This represents the predicted material flow rate after the total measured delay time. Represents an independent time variable; This indicates the real-time material flow rate at the current moment. This indicates the specific range of change in material flow rate within a preset, very short time period. This indicates the time span of the preset minimum time period; dividing the two gives the... This constitutes the instantaneous rate of change of the current material flow rate.

[0048] Using this formula, the central control unit can artificially shift the time axis forward, transforming lagging sensing data into advanced predictive data, thus providing a solid and accurate information foundation for subsequent feedforward control.

[0049] like Figure 2 The graph shown is a comparison curve of the effect of advance compensation for material flow data delay across devices. The horizontal axis of the graph represents the system running time, and the vertical axis represents the material flow.

[0050] Figure 2The actual flow rate curve drawn with a solid black line represents the most realistic physical flow rate fluctuation state when the upstream feed pump station just outputs the material. It can be seen that the waveform curve begins to show a clear upward trend of peaks around the fifth second, with the highest flow rate value reaching about 120.

[0051] Figure 2 The original lag flow curve, drawn with a blue dashed line, represents the data transmitted directly to the central control unit of the paving vessel through long-distance physical pipelines and network communication without any processing. It can be clearly observed that the waveform has been severely shifted to the right on the time axis, and its peak arrival time is about 3 seconds later than the actual physical peak. If this blue dashed line data is used directly for feedforward control, it will inevitably lead to a misalignment of the propeller's power delivery time, thereby exacerbating the hull sway.

[0052] and Figure 2 The predicted compensation flow curve drawn with a solid red line is the final result of the control system of this invention performing advance prediction compensation on the time axis in the underlying algorithm. The solid red line perfectly matches and restores the changing trend of the solid black line in terms of waveform trend and phase node, successfully pushing the lagging perception data forward to the actual occurrence time on the time axis.

[0053] The above multi-dimensional curve comparison and waveform transformation intuitively demonstrate that the delay pre-compensation technology provided by this invention can eliminate time misalignment conflicts caused by long-distance cross-device collaboration with extreme precision, providing a solid and error-free data foundation for the real-time issuance of subsequent feedforward control commands. This initially verifies the beneficial effects of this technology in overcoming system misjudgment and preventing over-compensation of the actuator.

[0054] Optionally, even after obtaining accurately compensated cross-device collaborative data, directly injecting it into the control model still carries the risk of oscillation. This is because large slurry pumps, during operation, are affected by impeller rotation, pipe cavitation, and uneven material concentration, resulting in a significant amount of nonlinear high-frequency pulsation in the material flow data. Therefore, the system also includes a flow disturbance processing module, used to perform pulsation-level filtering on the cross-device collaborative data before the two-stage adaptive decision module generates control commands. This pre-processing data smoothing mechanism effectively filters out high-frequency noise that has no impact on the ship's macroscopic motion, preventing unnecessary high-frequency wear on the actuators due to noise.

[0055] Specifically, the execution process of the pulsation-level filtering includes: dividing the pulsation intensity into a first pulsation level, a second pulsation level, and a third pulsation level based on the variance of the material flow rate, and calling a moving average filtering algorithm with a different window width for each pulsation level. In actual operation, the central control unit collects material flow rate sample points over a period of time in real time and calculates their statistical variance. When the variance is small and below the primary threshold, the system determines that it is currently in the first pulsation level. At this time, the fluid is extremely stable, and the system calls a moving average filtering algorithm with a narrow window width to maintain the real-time performance of the signal to the greatest extent. When the variance is between the primary threshold and the intermediate threshold, the system determines it to be in the second pulsation level, and calls a filtering algorithm with a medium window width for moderate smoothing. When the variance is greater than the intermediate threshold, it indicates that there may be an extreme pulse situation in the pipeline where a large piece of material has blocked the pipeline and then suddenly cleared the blockage. The system determines it to be in the third pulsation level, and calls a filtering algorithm with the widest window width to forcibly suppress sudden spikes with strong low-pass characteristics.

[0056] It's also important to note that adjusting the filter window solely based on variance has limitations, as the back pressure of the pipeline system is a core parameter reflecting the fluid state. Therefore, this process also includes dynamically updating the filter coefficients of the moving average filtering algorithm based on the differential pressure value of the material pipeline. The differential pressure value represents the pressure change trend. When the differential pressure value is positive and increases sharply, it indicates that the pipeline is experiencing rapid pressure buildup. Even if the flow variance is temporarily small, the system will immediately anticipate the impending strong pulsation and dynamically update the weight coefficients in the moving average filtering algorithm in advance, giving greater weight to historically stable data. This results in a smoother flow curve after filtering, effectively eliminating the root cause of system control divergence.

[0057] like Figure 3 The figure shows a comparison of the effects of material flow rate pulsation-based graded filtering based on pressure derivative adaptive adjustment. The horizontal axis of the figure represents the system running time, and the vertical axis represents the material flow rate value.

[0058] Figure 3 The raw material flow curve, drawn with a thin gray solid line, represents the unprocessed primary fluid data directly collected by the system from the cross-device collaboration end. It can be clearly observed that when the system runs for about ten seconds, due to abnormal conditions such as severe pressure buildup and release inside the pipeline, the curve instantly bursts with extremely destructive high-frequency pulse spikes, and its up and down oscillation amplitude is extremely violent.

[0059] Figure 3The conventional graded filtering curve drawn with a blue dashed line represents the filtering effect of existing technology that relies solely on the variance of the flow rate for passive post-adjustment. Due to the inevitable time lag in variance calculation and grade determination, the blue dashed line still exhibits a violent jump in the early stage of sudden pulses, failing to intercept destructive spikes in time.

[0060] In comparison, Figure 3 The adaptive adjustment filter curve, drawn with a thick red solid line, represents the filtering result after the control system of this invention proactively triggers a dynamic weight update when it detects a sharp increase in the differential value of the pipeline pressure. Faced with the violent high-frequency oscillations of the thin gray solid line, the thick red solid line not only shows no abrupt changes in response but also maintains an extremely smooth and stable transition trend.

[0061] The stark contrast of the waveforms above vividly demonstrates that the technical solution provided by this invention, which combines pressure change trends to dynamically update the pre-filter parameters, can overcome the passive lag limitation of single flow monitoring. Before severe operating conditions actually occur and cause violent flow fluctuations, it establishes a robust data smoothing barrier in advance, as if it has predictive capabilities. This completely eliminates the causes of control system divergence and high-frequency ineffective wear of the propulsion actuator from the source, and improves the robustness and command stability of the control system under extreme operating conditions.

[0062] For example, after data acquisition, delay compensation, and pulsation-level filtering are completed, the system's core processing center begins to intervene. This system includes a dynamic disturbance model construction module, used to construct a dynamic disturbance model based on the paving vessel's local data and the cross-equipment collaborative data after delay compensation, and to calculate and generate feedforward compensation amounts for the thrusters and anchor winches through this dynamic disturbance model. Traditional ship dynamic positioning systems often only establish environmental disturbance models based on wind, waves, and currents, while the dynamic disturbance model in this embodiment creatively incorporates the technological process disturbances of the paving operation itself into the overall calculation.

[0063] Specifically, the dynamic disturbance model construction module constructs the dynamic disturbance model by extracting characteristic variables from the local data and cross-equipment collaborative data of the paving vessel. These characteristic variables include at least material flow rate, material pump outlet pressure, material concentration, nozzle attitude, draft, ocean current velocity, and wave height. These seven characteristic variables constitute the full-dimensional vector space for the force analysis of the paving vessel. Material flow rate and material pump outlet pressure directly determine the enormous recoil thrust generated when material is ejected from the riprap nozzle. Material concentration determines the fluid density, thus affecting the absolute amplitude of the recoil force. Nozzle attitude determines the vector decomposition direction of the recoil force in three-dimensional space, altering the yaw moment and lateral force acting in the hull coordinate system. Changes in draft nonlinearly alter the wetted surface area of ​​the hull underwater, thereby fundamentally changing the hydrodynamic damping coefficient experienced by the hull. Ocean current velocity and wave height are rigid external disturbances in the marine environment, generating direct water flow thrust and wave radiation force.

[0064] To uniformly map these characteristic variables, which have completely different physical dimensions and mechanisms of action, to the forces acting on the hull, the process further includes: assigning corresponding characteristic weight coefficients to each of the characteristic variables. The characteristic weight coefficients are the core parameters for the system's mechanical equivalence transformation; essentially, they represent the contribution rate of a unit change in a specific characteristic variable to the overall torque of the hull.

[0065] For example, the process then proceeds as follows: each of the aforementioned feature variables is multiplied by its corresponding feature weight coefficient to obtain each feature component, and all the feature components are weighted and summed to calculate the disturbance moment characterizing the degree of disturbance to the paving vessel. Through this highly linear algebraic mapping mechanism, the central control unit can collapse the complex sensor data distributed throughout the ship and the external environment into an intuitive and clear physical quantity, namely the disturbance moment, within a one-microsecond calculation cycle. This disturbance moment is then fed into a dedicated feedforward control law calculation stage. After solving the inverse model of ship dynamics, the additional pre-force required by the side thrusters and anchor winches at the next moment to counteract this disturbance moment is accurately deduced; this is the aforementioned feedforward compensation amount.

[0066] It is also important to note that the marine environment and paving conditions are extremely variable. The fluid resistance characteristics are completely different in the initial and later stages of rock-laying; the frequency of hydrodynamic disturbances also varies significantly during high and low tides. If a fixed set of static characteristic weighting coefficients is used, the system will inevitably develop severe steady-state errors after running for a period of time. Therefore, the dynamic disturbance model construction module also includes a hybrid identification unit for online evolution of these parameters.

[0067] Specifically, the execution logic of the hybrid identification unit is as follows: A clustering algorithm is used to classify the disturbances experienced by the paving vessel into flow-dominated disturbances, sea state-dominated disturbances, and coupled disturbances. During the background calculation process, the system continuously collects all feature variables and their derived features within a certain time window. The clustering algorithm searches for data cluster centers in the high-dimensional data space. If the variance of material flow and pressure dominates in the current data sample cluster, while the ocean current and wave data are relatively calm, the clustering algorithm outputs a category label, classifying the current disturbance as a flow-dominated disturbance. Conversely, if the wave height changes drastically under the influence of the outer circulation of a typhoon while the material supply system remains in a stable standby state, it is classified as a sea state-dominated disturbance. If severe sea conditions occur simultaneously with full-load rock-dropping operations and simultaneous internal and external disturbances, the system classifies it as the most dangerous coupled disturbance.

[0068] like Figure 4 The figure shows the clustering spatial distribution and classification boundary division of multi-source disturbance characteristics under complex working conditions. The horizontal axis of the figure represents the comprehensive characteristic value of the internal pipeline fluid state change, and the vertical axis represents the comprehensive characteristic value of the external marine environment state change. The scale of the coordinate axis is normalized from 0 to 1. Figure 4 The paper demonstrates three typical perturbation feature scatter sets continuously collected and mapped into a high-dimensional data space by the hybrid identification unit in the background.

[0069] The set of red circular data points in the lower right corner represents the flow-dominated disturbance distribution data. The horizontal coordinate values ​​of these points are mostly concentrated in the higher range of around 0.75, while the vertical coordinate values ​​remain in the lower range of around 0.2. This intuitively reflects the physical condition where the variance of material flow and pipeline pressure dominates, while the ocean current and waves are relatively calm.

[0070] The set of blue circular data points in the upper left corner represents the distribution data of sea state-dominated disturbances. Its vertical axis value is relatively high while its horizontal axis value is relatively low, which corresponds to the scenario where the wave height changes drastically under the influence of the outer circulation of the typhoon while the feeding system is in a standby and stable state.

[0071] The set of purplish-red circular data points in the upper right corner represents the most dangerous coupled perturbation distribution data, where internal and external excitations occur simultaneously, and both horizontal and vertical characteristic values ​​are at high levels.

[0072] The three black pentagrams in the figure represent the three baseline centers of the clusters that the clustering algorithm finds and anchors in space, while the black dashed line running through the entire feature space is the absolute classification boundary line generated by the algorithm through the spatial geometric distance metric function.

[0073] It can be clearly observed that a large number of edge data points are scattered along the black dashed boundary areas where various categories intersect. When the marine construction environment changes continuously, the system's real-time feature data will fall precisely in these blurred boundary areas and experience slight fluctuations on both sides of the dashed lines. At this time, if traditional hard classification logic is used, it will inevitably lead to high-frequency jumps in control parameters that cross the boundary. However, this figure clearly reveals that the geometric distances from these boundary-confusing data points to the center points of the two adjacent black pentagram reference points are extremely close. This proves the absolute necessity and physical rationality of actively determining the transition state, activating the dual-track parallel processing mechanism, and extracting the reciprocal distance for cross-weighted fusion in this area.

[0074] Through this visualization of the clustering spatial distribution and boundary relationships, we can initially acknowledge the substantial advantages of the smooth transition update logic of this invention in resolving the problems of algorithm hard switching and parameter divergence collapse.

[0075] For example, after determining the current disturbance category, the hybrid identification unit triggers corresponding parameter update algorithms to iteratively update the feature weight coefficients in the dynamic disturbance model for different types of disturbances. The reason for triggering different parameter update algorithms is that different disturbances have drastically different spectral characteristics and convergence requirements. When facing high-frequency, pulsating flow-dominated disturbances, the system-triggered parameter update algorithm focuses on online identification logic with extremely fast forgetting speed to quickly capture transient changes in system characteristics. When facing periodically significant and low-frequency sea state-dominated disturbances, the system-triggered parameter update algorithm focuses on gradient descent optimization search to robustly correct the feature weight coefficients corresponding to draft and wave height based on long-term statistical regularities. Through this context-specific hybrid identification mechanism, the dynamic disturbance model gains strong vitality and adaptability, ensuring that the output disturbance torque always possesses extremely high physical realism regardless of how harsh or alternating the external environment.

[0076] After completing the precise calculation of the feedforward compensation and the self-iteration of the model, the system enters the final decision-making and control output stage. This system includes a two-stage adaptive decision module, used to generate control commands based on the feedforward compensation and feedback decoupling optimization logic to adjust the actual output of the thruster and the anchor winch. The two-stage structure here means that the system neither blindly trusts the feedforward commands nor simply relies on feedback correction, but rather establishes a comprehensive decision-making architecture that includes safety constraints and fusion ratios.

[0077] Specifically, the process of generating control commands by the two-stage adaptive decision module includes: decomposing the feedforward compensation amount into feedforward thrust and feedforward pull. Since the actuators of the paving vessel include not only the side thrusters that provide active force vector deflection but also the anchor winch system that provides passive mooring constraints, these two differ significantly in spatial layout and response time. Therefore, the system must utilize matrix mapping technology to decompose the unified, abstract feedforward compensation amount into feedforward thrust commands suitable for execution by the side thrusters, and feedforward pull commands suitable for the anchor winch to control cable deployment and retraction.

[0078] It is also important to note that a critical weakness of feedforward control is its extreme sensitivity to model errors. Once an abnormally sudden disturbance signal is input, the calculated feedforward quantity can easily cause instantaneous overload saturation of the actuator. Therefore, the module employs a strict firewall mechanism before issuing commands, specifically checking whether the feedforward thrust exceeds a preset thrust threshold and whether the feedforward pull exceeds a preset pull threshold. The preset thrust threshold is typically set to 85% of the rated power output of the thruster, thus reserving a power margin to handle feedback deviation corrections; the preset pull threshold is set as the critical value obtained by subtracting a safety factor from the safe breaking pull of the anchor winch wire rope.

[0079] For example, if the feedforward thrust is greater than a preset thrust threshold or the feedforward pull is greater than a preset pull threshold, it indicates that the predicted future disturbance has exceeded the hardware resistance limit of the paving vessel itself. Forcing execution at this point would inevitably lead to overcurrent tripping of the side thrusters or breakage of the anchor cable. Therefore, the system adopts a safety-first strategy: reducing the value of the feedforward compensation by a preset attenuation step and issuing a flow restriction command to the material pump. This cross-device linkage restriction mechanism is a core highlight of this system. By weakening the feedforward compensation by a step, the actuator is ensured to operate at full capacity within the maximum safety boundary; simultaneously, the material pump flow is forcibly reduced at the feeding end, eliminating the power source causing the flow-dominated disturbance from the root and quickly mitigating the risk of system collapse.

[0080] Conversely, if the feedforward thrust is not greater than a preset thrust threshold and the feedforward pull is not greater than a preset pull threshold, the system determines that the current operating condition is within a controllable range. Then, based on the currently allocated feedforward and feedback weights, the system weightedly fuses the feedforward thrust and the feedback-optimized thrust to generate the final thrust command, and weightedly fuses the feedforward pull and the feedback-optimized pull to generate the final pull command. During this process, the feedback-optimized thrust and feedback-optimized pull are calculated in real-time by the system's underlying classical proportional-integral-derivative controller based on the current pose error. The feedforward mechanism is responsible for proactively mitigating most of the coarse-grained disturbance trends, while the feedback mechanism is responsible for eliminating the minor steady-state errors caused by model mismatch. Through the weighted fusion of the two weights, a perfect unity of proactive prediction and delayed correction is achieved. The final generated control command can drive the side thrusters and anchor winch to smoothly, gently, and extremely precisely adjust the hull attitude and position.

[0081] Optionally, the feedforward and feedback weights mentioned above are not static, but rather require dynamic negotiation based on the ship's current energy consumption distribution and control precision. In large-scale marine engineering, the side thrusters, as high-power electrical equipment capable of full azimuth rotation, have extremely high energy costs. Therefore, the allocation process of feedforward and feedback weights in the two-stage adaptive decision-making module includes an extremely sophisticated energy consumption perception mechanism.

[0082] Specifically, the allocation process first involves calculating real-time energy consumption based on the actual output thrust of the side thrusters, the shifting speed of the paving vessel, the actual output pulling force of the anchor winch, and the anchor winch's deployment and take-up speed. The principle of work in physics states that power equals the dot product of the force and the velocity vector. The system calculates the mechanical power of the propulsion system by extracting the actual output thrust calculated from the output torque and rotational speed of the side thruster's frequency converter, combined with the shifting speed provided by the high-precision navigation system; simultaneously, it extracts the driving pulling force of the anchor winch's hydraulic motor and the cable deployment and take-up speed to calculate the damping power of the mooring system. Transient integration and filtering of these two parameters yield the system's precise real-time energy consumption at the current moment.

[0083] Subsequently, the system calculates the energy consumption deviation between the real-time energy consumption and the preset optimal energy consumption reference value. The preset optimal energy consumption reference value is a dynamic baseline derived from long-term self-learning under stable operating conditions, representing the minimum necessary energy consumption required to complete the current construction task. The magnitude of the energy consumption deviation directly reflects whether a large amount of energy is wasted in the high-frequency useless reciprocating adjustments of the actuators in the current control strategy.

[0084] For example, in response to the condition that the energy consumption deviation value is greater than a preset deviation threshold and the current pose error of the paving vessel is less than a preset error threshold, the system proportionally reduces the feedforward weight and increases the feedback weight to maintain a constant sum of the feedforward weight and the feedback weight. This concurrency condition demonstrates remarkable engineering ingenuity. An energy consumption deviation value greater than the preset deviation threshold indicates that the current system is extremely power-intensive, and the side thrusters may be frequently accelerating and decelerating. Simultaneously, a current pose error less than the preset error threshold indicates that although the hull control is very precise, this high precision comes at the cost of excessive energy consumption and severe mechanical wear.

[0085] Because feedforward control is extremely sensitive to minute high-frequency disturbances and is the culprit behind this high-frequency wasted effort, the system decisively decides to proportionally reduce the feedforward weight. As the feedforward channel is weakened, the high-frequency commands that cause frequent flutter in the thrusters are significantly suppressed; simultaneously, the feedback weight is increased, relying on the integral lag characteristic of the feedback loop to maintain the macroscopic stability of the hull. While ensuring that the sum of the two remains constant, the entire control system achieves a smooth transition from an aggressive, high-energy-consuming tracking mode to a conservative, energy-saving cruise mode. This not only extends the service life of expensive marine equipment but also significantly reduces the unit reclamation cost.

[0086] like Figure 5 The figure shows the dynamic game adjustment curve of feedforward and feedback weights in a two-order adaptive control system based on energy consumption perception. The horizontal axis of the figure represents the system running time, the left vertical axis represents the weight value, and the right vertical axis represents the real-time energy consumption value.

[0087] The feedforward weight curve drawn with a solid red line and the feedback weight curve drawn with a dashed blue line in the figure show a clear mirror-symmetric game relationship, and the sum of their values ​​is always strictly kept to be one.

[0088] In the initial 15 seconds of system operation, the paving vessel's attitude error is relatively large, and the system is in an aggressive, high-precision tracking mode. The feedforward weight (red solid line) remains at a high level of 0.6, while the feedback weight (blue dashed line) remains at a low level of 0.4. Observing the real-time energy consumption curve plotted as a black solid line in the figure, it can be seen that during this stage, because the feedforward control is extremely sensitive to small high-frequency disturbances, the thrusters frequently accelerate and decelerate, and the real-time energy consumption curve remains at a high level of around 80, accompanied by violent high-frequency oscillations.

[0089] When the system reaches the fifteenth second, the current pose error of the paving vessel has gradually converged and is less than the preset error threshold, but the real-time energy consumption remains high and its deviation is greater than the preset deviation threshold. At this time, the system very sensitively triggers the energy consumption perception weight allocation mechanism.

[0090] Within the dynamic game adjustment period from the 15th to the 25th second, the red solid line representing the feedforward channel begins to show an extremely smooth downward trend and eventually stabilizes at a low level of 0.2, while the blue dashed line representing the feedback loop rises proportionally and stabilizes at a high level of 0.8. As the feedforward weight is passively weakened and the feedback weight gradually takes over, the high-frequency commands that cause frequent oscillations are significantly suppressed. From the intuitive physical effect reflected by the black solid line, the real-time energy consumption curve immediately shows a precipitous smooth downward trend after entering the adjustment period, and finally stabilizes at an extremely low energy consumption level of around 40 after the 25th second, with extremely small oscillation amplitude.

[0091] The stark color contrast of the curves, the mirrored intersection of their trends, and the significant decrease in energy consumption figures powerfully demonstrate that the dual-stage adaptive decision-making module provided by this invention can smoothly achieve the intelligent transition of the system from an aggressive, high-energy-consuming tracking mode to a conservative, energy-saving cruising mode while ensuring the macroscopic positional stability of the hull. This significantly reduces the ineffective energy consumption throughout the underwater construction process and fundamentally avoids severe mechanical wear of core marine engineering equipment caused by high-frequency flutter.

[0092] Finally, it is crucial to remember that safety during marine construction is always the bottom line. This system not only pursues ultimate control precision and economy but also establishes a redundant protection network designed to cope with catastrophic extreme conditions. The system also includes a multi-dimensional anomaly handling module to monitor and intercept systemic risks that individual modules cannot handle.

[0093] Specifically, the multi-dimensional anomaly handling module monitors system operating parameters in real time. In response to concurrent conditions where the current material flow rate change rate exceeds a preset flow mutation threshold, the current sea state level is greater than or equal to a preset danger level threshold, and the current pose error growth rate exceeds a preset displacement growth rate threshold, a coupled anomaly signal is generated. The concurrence of these three conditions represents the most extreme working scenario encountered by the paving vessel: a severe pulsed water hammer effect may have occurred within the pipeline, causing a rapid change in flow rate; simultaneously, it is being battered by the gale-force winds and waves of a typhoon front; and the original control commands of the system have completely failed, with the hull deviating at an extremely rapid acceleration. At this point, conventional feedforward and feedback adjustments are not only ineffective but may even exacerbate the risk of capsizing.

[0094] Faced with this extremely urgent situation, the multi-dimensional anomaly handling module has the highest execution priority within the system. In response to the coupled anomaly signal, it triggers a cross-device coordinated halt action and issues a mandatory emergency control command to the two-stage adaptive decision module. The cross-device coordinated halt action directly sends a hard-wired emergency stop signal to the feed vessel, instantly cutting off the main power supply to the material pumps and closing all sludge discharge main valves, blocking all sources of internal human disturbance. Simultaneously, the mandatory emergency control command bypasses complex model calculations, instructing the two-stage adaptive decision module to immediately lock all current state variables, drive all side thrusters into a typhoon-resistant positioning mode with maximum power to resist external wind and waves, and control the anchor winch to quickly tighten all fixed-point cables, ensuring the vessel's attitude does not deteriorate further. This buys valuable time for on-site personnel to intervene and expedite the rescue, raising the system's safety to the highest standard.

[0095] In summary, the adaptive displacement control system for pavers based on flow feedforward provided in this embodiment lays a high-precision information source foundation through cross-device delay compensation preprocessing, uses a dynamic feature weight matrix to reduce the dimension of multi-source coupled physical quantities and map them into precise disturbance torques, achieves adaptive convergence under complex working conditions through clustering identification and pulsation hierarchical filtering, and forms an intelligent whole with full-cycle closed-loop capabilities of perception, prediction, optimization, and self-healing through energy consumption perception feedforward feedback two-order dynamic matching and multi-dimensional anomaly fallback mechanism. This fills the technical gap in the field of collaborative micro-motion control of large special engineering vessels under harsh sea conditions.

[0096] Example 2:

[0097] This embodiment is based on the system architecture of Embodiment 1, and further optimizes and expands the technical features to address the strong coupling technical challenges induced by the overlap of extremely harsh marine meteorological environments and complex construction operation conditions.

[0098] In conventional paving operations, wave undulations and material pump pressure changes are typically considered two independent disturbance sources, and the linear weighted summation model in Example 1 can perfectly handle the superposition of these independent disturbances. However, in actual severe sea conditions, when sudden strong winds and giant waves impact the paving vessel, the hull experiences violent pitching and heaving motions. This drastic change in spatial position causes a sudden and rapid change in the depth of the underwater rock-filled pipe outlet. This rapid change in outlet depth directly alters the hydrostatic pressure at the pipe's end, instantly transmitting this enormous reverse water pressure to the upstream material pump, resulting in a violent and abnormal pulse in the pump's outlet pressure.

[0099] Under these extreme conditions, external environmental waves and internal pipeline pressure are no longer two independent physical quantities, but rather form a strongly coupled relationship of mutual cause and effect. If the control system continues to mechanically calculate and directly weight and sum these two abrupt characteristic variables based on the same physical cause, the central control unit will make a serious misjudgment, mistakenly believing that the paving vessel has suffered a double fatal blow from both external giant waves and severe internal pipeline blockage. This misjudgment will directly lead the system to calculate an abnormally large disturbance torque, thereby generating an extremely aggressive feedforward compensation, causing the side thrusters to instantly operate at full load or even overload at full speed in the opposite direction. This not only completely disrupts the predetermined trajectory of the paving operation, but also causes violent reverse oscillations of the hull, resulting in a serious overcompensation disaster.

[0100] To completely resolve this conflict-related technical challenge, the dynamic disturbance model construction module also includes an extreme coupling decoupling unit. This extreme coupling decoupling unit, serving as the core of the system's safety in responding to extreme sea conditions, can identify and eliminate this spurious disturbance superposition effect at the underlying logic level.

[0101] Specifically, the core pre-judgment logic executed by the aforementioned extreme coupling decoupling unit is to monitor the wave height change rate and the material pump outlet pressure change rate in real time. To achieve millisecond-level response to sudden severe operating conditions, the extreme coupling decoupling unit does not rely on macroscopic average numerical analysis, but focuses on the transient differential characteristics of high-frequency sampling. The system continuously sends the current absolute wave height and absolute material pump outlet pressure data into the buffer register of the central control unit using a high-frequency multibeam wave meter and a high-precision pressure transmitter.

[0102] The wave rate of change is calculated using the following formula:

[0103] ;

[0104] In the formula, The rate of change of wave height; This indicates the real-time wave height at the current sampling moment; This represents the real-time wave height at the previous sampling time. This represents the time interval between two adjacent sampling times.

[0105] The pressure change rate is calculated using the following formula:

[0106] ;

[0107] In the formula, The rate of change of pressure at the outlet pressure of the material pump; This indicates the real-time material pump outlet pressure at the current sampling moment; This indicates the real-time material pump outlet pressure at the previous sampling time.

[0108] Through the two rigorous micro-discretion equations mentioned above, the extreme coupling decoupling unit can transform static amplitude data into dynamic trend data, and extremely sensitively capture any minute signs of abrupt changes in the sea surface environment and inside the pipeline.

[0109] It is also important to note that not all changes require intervention; the paving vessel's inherent inertia can automatically filter out minor fluctuations. Therefore, the extreme coupling-decoupling unit has an extremely strict trigger threshold.

[0110] The system responds to the operating conditions where the wave change rate exceeds a preset wave mutation threshold and the pressure change rate exceeds a preset pressure mutation threshold, triggering an overcompensation suppression strategy. This concurrent logic judgment is an absolute prerequisite for the entire decoupling mechanism to activate. The preset wave mutation threshold is a critical value pre-calibrated through simulation calculations based on the paving vessel's stability arm curve and its wind and wave resistance design level. It represents the limit of wave rise speed sufficient to cause irresistible displacement of the hull. The preset pressure mutation threshold is a critical value calibrated based on the flow-pressure characteristic curve of the material pump and the pipeline's safe burst pressure.

[0111] Only when the wave height experiences a catastrophic surge in an instant, and the material pump outlet pressure also spikes dramatically within the same extremely short period, will the system definitively determine that it is currently experiencing a strongly coupled extreme operating condition involving internal and external variables. This dual-abrupt concurrency condition setting effectively avoids the mis-triggering of decoupling logic caused by local turbulence or noise from a single device under normal sea conditions, ensuring the normal stability of the control system. Once the aforementioned concurrency condition is met, the system's main thread calculation will be immediately interrupted, and the system will be awakened with the highest priority and trigger the overcompensation suppression strategy to avert disaster and salvage the impending loss of control over the thrust calculation process.

[0112] In the overcompensation suppression strategy, the primary task of the system is to quantify the degree of this strong coupling. To this end, the system extracts a first data sequence of wave height and a second data sequence of material pump outlet pressure within a preset historical time window. The preset historical time window is a dynamically sliding memory period, typically set to contain at least two complete wave cycles. The system precisely extracts all continuous sampling points from the circular buffer of the underlying database, counting backwards from the trigger moment towards the preset historical time window. All extracted wave height sampling points are arranged in chronological order to form the first data sequence. Simultaneously extracted material pump outlet pressure sampling points are arranged with identical time labels to form the second data sequence. These two data sequences maintain absolute millisecond-level alignment on the time axis, thus providing a precise set of raw data samples for subsequent rigorous statistical mathematical analysis.

[0113] Specifically, after acquiring the aligned data sequences, the extreme coupling decoupling unit immediately executes and calculates the Pearson correlation coefficient between the two data sequences. The Pearson correlation coefficient is a powerful mathematical tool in multivariate statistical analysis for accurately measuring the strength of a linear correlation between two continuous variables.

[0114] The Pearson correlation coefficient is calculated using the following formula:

[0115] ;

[0116] In the formula, This represents the Pearson correlation coefficient; This indicates the total number of discrete sampling data points extracted within the preset historical time window; This represents the index ordinal number of the discrete sampled data points, with a value ranging from 1 to... Positive integers; Represents the first data sequence in which the first data sequence is... The specific value of each wave height is sampled. This represents the global arithmetic mean of all wave height samples in the first data sequence; Indicates the first in the second data sequence Specific values ​​of the outlet pressure of each material pump were sampled. This represents the global arithmetic mean of all sampled values ​​of the material pump outlet pressure in the second data sequence.

[0117] By calculating the sum of the products of the deviations of each instantaneous value from its sequence mean, and then dividing by the product of the square roots of the sum of the squares of the respective deviations, the system extremely accurately eliminates the interference of absolute amplitude magnitude on correlation judgment, directly revealing the degree of synchronous resonance between wave fluctuations and pressure pulses in the time dimension. If the Pearson correlation coefficient is extremely close to the value of one, it fully proves that the current huge pressure pulse is entirely a passive physical phenomenon caused by the backflow of water pressure from external strong winds and waves through the rock-throwing pipe, rather than a true pump source power anomaly.

[0118] For example, after accurately quantifying the degree of coupling, the extreme coupling decoupling unit calculates the attenuation ratio of the characteristic weight coefficient for the material pump outlet pressure based on the Pearson correlation coefficient, where the value of the Pearson correlation coefficient is positively correlated with the magnitude of the attenuation ratio. This step represents a leapfrog transformation from a purely statistical indicator to a specific engineering control parameter. In Example 1, the characteristic weight coefficient is defined as a multiplier factor measuring the contribution of a specific characteristic variable to the hull disturbance torque. After determining it to be a strong coupling pseudo-disturbance superposition, this pseudo-contribution must be forcibly weakened.

[0119] The attenuation ratio is calculated using the following formula:

[0120] ;

[0121] In the formula, Indicates the attenuation ratio; This represents the preset attenuation adjustment constant; This represents the previously calculated Pearson correlation coefficient.

[0122] The preset attenuation adjustment constant is an empirically calibrated fixed value used to match the overall gain of the control system. The rigorous calculation logic described above ensures an absolute positive correlation between the Pearson correlation coefficient and the magnitude of the attenuation ratio. This means that the more pronounced the synchronized effect of wave and pressure, the more false components are identified in the pressure abrupt change, and the larger the attenuation ratio calculated by the system, thus issuing a more severe deweighting penalty.

[0123] It is also important to note that the final step of the extreme coupling decoupling unit is to perform substantial model parameter intervention. The system uses the attenuation ratio to update the characteristic weight coefficients of the material pump outlet pressure with reduced weights.

[0124] Here, the weighted update formula is used for calculation:

[0125] ;

[0126] In the formula, This represents the feature weighting coefficient for the material pump outlet pressure after the weight reduction update. This represents the original feature weighting coefficients for the material pump outlet pressure maintained by the system before the compensation suppression strategy was triggered; This represents the attenuation ratio calculated based on the degree of strong coupling. Through this mathematical substitution process, the original feature weight coefficients, which are prone to causing huge calculation errors, are proportionally and forcibly compressed into an extremely small or even close to zero value.

[0127] After completing this crucial underlying weight modification operation, the extreme coupling decoupling unit returns control to the main control thread and recalculates the disturbance torque using the updated feature weight coefficients. At this point, because the feature weight coefficient of the material pump outlet pressure has been significantly reduced, the pseudo-torque component contributed by the pipeline pressure is almost perfectly eliminated when the dynamic disturbance model performs a new multivariate weighted summation. The disturbance torque ultimately calculated by the system truly and purely reflects the substantial environmental force impact of severe sea conditions on the hull.

[0128] Based on this deeply decoupled and purified real disturbance torque, the feedforward compensation subsequently generated by the two-order adaptive decision module becomes extremely rational and stable. The final thrust command received by the side thrusters no longer contains those crazy overcompensation components caused by dependent variable coupling, thus completely avoiding the catastrophic risks of thrust oscillation and loss of hull control.

[0129] like Figure 6 The figure shows a comparison of the thruster control command output before and after decoupling from the strong coupling state of waves and pressure under extremely severe sea conditions. The horizontal axis of the figure represents the system running time, and the vertical axis represents the thruster thrust command value.

[0130] Figure 6 The undecoupled thrust command curve, drawn with a red dashed line, represents the control effect when using the traditional linear superposition model. It can be clearly observed that when the system encounters a sudden gale and huge wave impact around the tenth second of operation, the external wave fluctuations and the sudden change in the hydrostatic pressure of the internal pipeline form a strong physical causal coupling. The traditional algorithm treats the two as independent disturbances and incorrectly superimposes them, causing the red dashed line to erupt with an extremely distorted and violent overshoot peak in the interval from the tenth to the twenty-fifth second. Its thrust command value instantly soars to the dangerous extreme value of full load and is accompanied by violent high-frequency oscillations. This erroneous overcompensation command will inevitably cause the hull to sway violently in the opposite direction or even deviate from the intended operating trajectory.

[0131] In comparison, Figure 6The decoupled thrust command curve, drawn with a thick blue solid line, represents the control effect after introducing the extreme coupling decoupling unit in this invention. When the underlying algorithm accurately identifies this strongly coupled pseudo-disturbance through dual differential mutation monitoring and calculation of the Pearson correlation coefficient of historical sequences, the system adaptively generates a large attenuation ratio and forcibly reduces the weight of the characteristic weight of the material pump outlet pressure. Reflected in the waveform trend, when facing the same extreme sea conditions, the thick blue solid line eliminates the spurious torque component caused by pipeline pressure backflow; its waveform only exhibits the smooth upward trend required to cope with real environmental forces, and the maximum thrust command value is extremely stably limited to a reasonable and safe range of approximately 85.

[0132] The stark contrast and sharp comparison between the two curves under abrupt change conditions demonstrate that the dynamic weighting and decoupling technology based on sequence correlation quantification provided by this invention can fundamentally eliminate control distortion conflicts caused by strong coupling of multiple variables. This ensures that the paving vessel can still output extremely rational and stable displacement control commands under extremely severe sea conditions such as the periphery of typhoons, thereby avoiding the devastating risks of thrust oscillation and loss of hull control, and building a solid safety line for high-end underwater construction equipment.

[0133] In summary, this embodiment details how, in complex marine engineering applications, the system creatively resolves the strong coupling conflict between the external wave environment and the internal pipeline fluid dynamics through a coherent logical chain of dual differential mutation monitoring, historical sequence sliding extraction, Pearson correlation depth quantification, and dynamic weighted feedback. This mechanism not only plugs the logical loopholes of the multivariate linear superposition model at its theoretical source but also, at the engineering practice level, provides the paving vessel with a robust armor for continuous high-precision and high-stability relocation operations in extremely harsh sea conditions such as the periphery of typhoons, significantly improving the safety boundary and adaptive optimization capability of high-end underwater construction equipment.

[0134] Example 3:

[0135] This embodiment is based on the aforementioned embodiment and addresses the problem of classification boundary jumps and control parameter divergence and collapse that occurs when the hybrid identification unit in the dynamic disturbance model construction module faces complex and continuously changing marine construction environments. It further reconstructs the underlying algorithm and expands the deep technical solution.

[0136] In real marine construction environments, the evolution of various environmental parameters and paving condition parameters is highly continuous, nonlinear, and filled with high-dimensional noise. The hybrid identification unit uses a clustering algorithm to classify the disturbances experienced by the paving vessel into different types and triggers the corresponding parameter update algorithm.

[0137] Specifically, when the paving vessel system is in the transitional phase where the flow pulsation gradually subsides and the external wave surge gradually intensifies, the system's real-time disturbance characteristic data will not experience a precipitous change. Instead, it will fall precisely within the multidimensional spatial fuzzy boundary zone between the two pre-defined classification clusters of flow-dominated disturbance and sea state-dominated disturbance. If the control module still adopts the traditional, absolute, hard-line classification logic of either / or, when the characteristic variable data experiences extremely small numerical fluctuations at this fuzzy boundary, the classification result of the clustering algorithm will undergo high-frequency jumps and interchanges within two adjacent extremely short sampling calculation cycles.

[0138] It is also important to note that such high-frequency jumps in clustering results are extremely destructive. This causes the central control unit to erratically switch between two parameter update algorithms with completely different underlying mathematical logics, resulting in discontinuous step changes in the preset feature weight coefficients in the dynamic disturbance model. These step changes not only cause the system to lose convergence but also cause the side thrusters to receive wildly jittering shift commands, severely jeopardizing their mechanical lifespan. To completely solve the technical challenge of control oscillations caused by this hard algorithm switching, the hybrid identification unit is equipped with smooth transition update logic. The introduction of this smooth transition update logic establishes a buffer zone for a soft landing of the algorithm within the system.

[0139] For example, the execution process of the smooth transition update logic first includes: when performing classification calculations using a clustering algorithm, calculating the distance values ​​from the current feature variable to the centers of each preset classification cluster, and extracting the first distance value with the smallest value and the second smallest distance value. The current feature variable is a multi-dimensional, real-time dynamic vector composed of multiple dimensions such as material flow rate, material pump outlet pressure, ocean current velocity, and wave height. Each preset classification cluster center represents a typical, pure, single disturbance type as a reference anchor point in the multi-dimensional data space. The system uses a precise spatial geometric distance metric function to quantify in real time the similarity between the current system operating characteristics in a multiple superposition state and these reference anchor points. The shorter the distance, the more closely the current system characteristics match the disturbance type represented by the reference anchor point.

[0140] Specifically, in order to perform the above spatial distance measurement process accurately, the hybrid identification unit uses the multidimensional Euclidean distance formula to perform numerical calculations and differentiation.

[0141] The distance value is calculated using the following formula:

[0142] ;

[0143] In the formula, This represents the Euclidean distance from the current feature variable to the center of the k-th preset classification cluster. This indicates the total number of feature variable dimensions defined by the system; The index ordinal number represents the dimension of the feature variable, and its value ranges from 1 to... Positive integers; This represents the real-time normalized collected value of the current feature variable in the j-th physical dimension; This represents the fixed constant coordinate value of the center of the k-th preset category cluster in the j-th physical dimension. Through the systematic traversal calculation of the above formula, the hybrid identification unit can obtain a set of numbers containing the distance values ​​of all preset categories. Subsequently, the system performs a quick sorting operation on this set of numbers, directly selecting the two top-ranking key values, which accurately yields the aforementioned first distance value with the smallest value and the second smallest distance value.

[0144] Optionally, after successfully extracting the two core distance values, the system needs to quantify the degree of ambiguity in which the current data point is located in the boundary zone. Therefore, the smooth transition update logic further executes the following steps:

[0145] Calculate the ratio of the first distance value to the second distance value, and in response to the ratio being greater than a preset boundary confusion threshold, determine that the current system is in a disturbance type transition state.

[0146] The ratio is calculated using the following formula:

[0147] ;

[0148] In the formula, Indicates the distance ratio; This represents the first distance value with the smallest value extracted after sorting. This represents the second smallest distance value extracted after sorting. Since the first distance value is mathematically and logically necessarily less than or equal to the second distance value, the range of this distance ratio is firmly limited to a value greater than 0 and less than or equal to 1.

[0149] It's also important to note that the preset boundary confusion threshold is a precise empirical value that infinitely approaches 1. When the distance ratio calculated by the central computing module exceeds this preset boundary confusion threshold, it means that the first and second distance values ​​are extremely close in absolute terms. From a physical perspective of the multidimensional geometric feature space, this indicates that the current real-time feature vector of the paving vessel is precisely hovering on the boundary line between the two core preset classification clusters. The system cannot definitively classify it as either a flow-dominated disturbance or a sea state-dominated disturbance. Faced with this ambiguous situation, the system decisively cuts off the absolute control of a single algorithm, formally determining that the system is currently in a disturbance type transition state. This proactively activates a dual-track parallel processing mechanism to prevent any single algorithm from causing divergence in model weight calculation results due to incomplete information intake.

[0150] For example, to ensure the smooth evolution of model feature parameters in the boundary region, the system performs the most crucial parallel solution operation:

[0151] During the transition state of the disturbance type, a first parameter update algorithm associated with the first distance value and a second parameter update algorithm associated with the second distance value are simultaneously activated to calculate the first and second model coefficient update amounts, respectively. The first parameter update algorithm represents the expert optimization logic corresponding to the disturbance type closest to the current state, while the second parameter update algorithm represents another rapidly emerging and significant disturbance type's expert optimization logic. The central control unit allocates completely independent memory computing threads for these two algorithms in the background processor, allowing them to deduce a set of feature weight coefficient adjustment and upgrade strategies based on the exact same current feature variable input state, thereby generating their own independent first and second model coefficient update amounts with extremely high efficiency.

[0152] Specifically, after successfully obtaining the update values ​​of two model coefficients that are inherently competing, the system must merge them into one through a compromise mechanism that conforms to objective physical laws to avoid instruction conflicts. Therefore, the smooth transition update logic executes the fusion instruction:

[0153] The reciprocal of the first distance value is extracted as the first fusion weight, and the reciprocal of the second distance value is extracted as the second fusion weight. The first and second fusion weights are then used to weight and fuse the update values ​​of the first and second model coefficients to generate the transitional update value. Since a smaller spatial distance value represents a higher feature similarity, and the corresponding independent parameter update algorithm has greater reference value, the system cleverly employs the mathematical reciprocal rule to perfectly reverse the disadvantage of geometric distance values ​​into an advantage in the final decision weight.

[0154] In order to ensure that the above weighted fusion operation meets the strict requirements of conservation and unbiasedness at the mathematical level, the hybrid identification unit performs standard normalization on the two inverse weights before performing the final fusion.

[0155] The transitional update amount is calculated using the following normalized weighted fusion formula:

[0156] ;

[0157] In the formula, This represents the final weighted transitional update amount generated by the system. and These still precisely represent the first distance value with the smallest numerical value and the second distance value with the second smallest numerical value, respectively. In the above fusion formula... That is, the mathematical prototype corresponding to the first fusion weight mentioned above. That is, the mathematical prototype corresponding to the aforementioned second fusion weight; This represents the update amount of the first model coefficients calculated independently by the first parameter update algorithm; This represents the amount of the second model coefficient update calculated independently by the second parameter update algorithm.

[0158] Through this extremely rigorous and logically self-consistent mathematical cross-integration model, the system achieves a smooth transition from quantitative accumulation to qualitative leap in model compensation parameters, completely eliminating the abrupt instruction jumps that any single control algorithm might bring during the handover period.

[0159] like Figure 7 The figure shown is a comparison of the smoothness of soft and hard switching of model feature weight parameters during the transition period of disturbance type. The horizontal axis of the figure represents the system running time, and the vertical axis represents the value of model feature weight parameters.

[0160] Figure 7 The traditional hard-switching weight parameter curve, plotted as a blue dashed line, represents the control effect of relying solely on absolute classification labels for algorithm iteration. It is clearly observed that during the perturbation type transition period from the 10th to the 20th second of system operation, the clustering classification results experience frequent back-and-forth jumps due to slight fluctuations in the real-time feature data at the fuzzy boundary between the two preset classification clusters. This either-or hard judgment causes the blue dashed line to exhibit extremely dense step-like abrupt changes and violent oscillations within the numerical range of 0.2 to 0.8. If such weight parameters with severe step spikes are directly injected into the dynamic perturbation model for torque reconstruction, it will inevitably cause the thruster to receive wildly jittering shift commands, thus seriously endangering the mechanical life of the equipment.

[0161] In comparison, Figure 7 The soft handover transition weight parameter curve, drawn with a thick red solid line, represents the control effect after introducing smooth transition update logic in this invention. When the system calculates and finds that the ratio of the first distance value to the second distance value is greater than the preset boundary confusion threshold, thus determining that it has entered the transition state, it decisively and proactively activates the dual-track parallel processing mechanism and uses the reciprocal of the spatial distance as the cross-fusion weight to weight and fuse the update amounts output by the two independent algorithms. Reflected in the waveform trend, the thick red solid line, within the same transition interval from the tenth to the twentieth second, smooths out the abrupt instruction steps caused by the alternation of a single algorithm, presenting an extremely smooth and gentle rising transition curve, and finally smoothly transitions from 0.2 to a new steady-state value of 0.8.

[0162] The strong contrast in smoothness between the two curves during the transition period vividly and intuitively demonstrates that the dual-algorithm soft landing fusion mechanism based on spatial cluster center distance evaluation, which is unique in this invention, can solve the problem of high-frequency jumps in the underlying identification algorithm under complex alternating conditions, realize seamless and smooth handover between different control strategies, and thus improve the operational stability of large marine engineering equipment and the adaptive convergence capability of control parameters in extreme environments.

[0163] It should also be noted that the final closing action of the smooth transition update logic is to perform a substantial model parameter replacement loop: using the transition update amount to iteratively update the preset feature weight coefficients in the dynamic perturbation model.

[0164] The final iteration of the model weights is completed using the following formula:

[0165] ;

[0166] In the formula, This represents the latest version of the preset feature weight coefficient matrix obtained after a smooth iterative update during the transition period. This represents the old version of the preset feature weight coefficient matrix left over from the previous operation execution cycle; This represents the high-precision transition period update amount just calculated in the previous step.

[0167] Through the aforementioned complete and logically rigorous smooth transition update logic architecture, the paving vessel adaptive displacement control system in this embodiment enables the dynamic disturbance model to avoid the neurotic algorithm jumps and violent coefficient oscillations that occur in traditional ship dynamic positioning systems when facing exceptionally complex and harsh marine construction cross-conditions. Even if the dominant environmental disturbance factor slowly and continuously shifts from the material pipeline end to the external wave end, the system can perform extremely smooth and seamless handover calculations between the two response strategies. After the updated preset feature weight coefficients are re-injected into the dynamic disturbance model, the calculated disturbance torque representing external disturbances will always maintain extremely high continuous-time characteristics and first-order derivative smoothness characteristics, thereby ensuring that the feedforward thrust and feedforward pull commands output by the two-order adaptive decision module are free of any glitches or abrupt changes. This underlying algorithm reconstruction greatly protects the physical safety of the underwater propulsion motor and the hydraulic pipeline of the heavy anchor winch, eliminating the last stability hazard of the multi-device collaborative control system in extreme engineering applications.

[0168] Example 4:

[0169] like Figure 8 As shown, this embodiment provides an adaptive displacement control method for pavers based on flow feedforward. The method provided in this embodiment is a macroscopic distillation of the operation steps of the aforementioned control system in actual underwater engineering paving operations.

[0170] With the continuous development of underwater engineering technology, paving vessels play a crucial role in high-precision underwater operations such as laying crushed stone foundations for cross-sea immersed tunnels. However, the offshore operating environment is extremely harsh and complex. Traditional control schemes, relying on single hysteresis feedback regulation or simple ideal state feedforward compensation, are prone to control conflicts involving strong variable coupling and overcompensation when faced with the superposition of multiple internal and external dynamic disturbances. To fundamentally overcome the problems of feedforward command timing misalignment and severe reverse oscillations of the hull caused by system misjudgment under extreme sea conditions in existing technologies, this embodiment discloses a complete method flow.

[0171] Specifically, an adaptive shift control method for pavers based on flow feedforward is applied to the central control unit of the paver, and includes the following core execution steps:

[0172] First, the central control unit executes the following steps: In response to the paving vessel's relocation command, it collects local data from the paving vessel and cross-device collaborative data, and performs delay compensation on the material flow data in the cross-device collaborative data. In actual construction operations, after responding to a high-precision relocation command, the paving vessel needs to immediately capture the status of various sensors distributed throughout the vessel and the downstream supply vessel via the industrial fieldbus network. The paving vessel's local data covers spatial physical quantities such as hull attitude, draft, and actuator status, while the cross-device collaborative data includes conveying status quantities such as material flow and pressure from the upstream pumping station. Due to the extremely long physical pipelines across devices and the time-consuming parsing of communication protocols, the material flow data received by the central control unit naturally has a significant physical transmission and communication delay. If this delay is not processed, it will inevitably lead to severe timing misalignment in the feedforward control.

[0173] It is also important to note that, in order to completely resolve the control lag conflict caused by the aforementioned time misalignment, the system must perform forward prediction of the time axis at the algorithm level. Specifically, the calculation process for delay compensation of material flow data includes two rigorous mathematical derivation sub-steps.

[0174] The first sub-step is to obtain the preset material pipeline length, real-time material flow rate, and sensor inherent delay, and calculate the total measurement delay. In this step, the method instruction requires the system to accurately unify the macroscopic fluid dynamic delay with the microscopic electronic communication delay to obtain a time constant characterizing the overall lag characteristics of the current feeding system.

[0175] The aforementioned delay compensation calculation process further includes a second sub-step: based on the current material flow rate change rate and the total measurement delay, calculate the predicted material flow rate at the actual moment of the disturbance, and input the predicted material flow rate into the dynamic disturbance model. By calculating the differential change trend of the fluid over a short period of time and combining it with the total delay time constant for linear extrapolation, the central control unit can artificially transform the lagging sensing data into the leading predictive data, which provides a solid and time-error-free information source foundation for subsequent precise feedforward control.

[0176] After completing the preliminary data synchronization and delay compensation, the central control unit continues to execute the next step: based on the local data of the paving vessel and the cross-equipment collaborative data after delay compensation, a dynamic disturbance model is constructed, and feedforward compensation amounts for the thrusters and anchor winches are calculated and generated through the dynamic disturbance model.

[0177] To address the control conflict caused by strong coupling of multiple variables under extreme conditions, as mentioned in the background section, this dynamic disturbance model does not simply perform a linear, independent superposition of all variables. During model construction, the system monitors in real-time the abrupt changes and sequence correlations of the core external forces and internal states. When encountering sudden strong winds and waves that cause a sudden change in the ship's draft and trigger a severe pressure pulse in the pipelines, the method adaptively calculates the attenuation ratio and dynamically reduces the weight of relevant feature coefficients, or utilizes a dual-algorithm soft-switching transition logic to cleverly and accurately eliminate the pseudo-disturbance superposition caused by strong physical coupling. The purified feature variables are then mapped to highly faithful disturbance torques, and ultimately, the feedforward compensation required to counteract the real disturbance is derived.

[0178] After accurately acquiring the feedforward compensation amount, the central control unit executes the final control decision and execution step: based on the feedforward compensation amount and feedback decoupling optimization logic, it generates control commands to adjust the actual outputs of the side thrusters and the anchor winch, and executes the adaptive displacement of the paving vessel. In this step, the system does not blindly issue feedforward commands, but combines the two-order adaptive decision-making mechanism described in Example 1. The system dynamically balances and proportionally integrates aggressive feedforward weights and conservative feedback weights based on the current energy consumption distribution and spatial pose error of the paving vessel. This feedback decoupling optimization logic ensures that the feedforward prediction mechanism can reduce most of the coarse disturbance trends in advance, while retaining the closed-loop correction capability of the underlying proportional-integral-derivative controller.

[0179] By fully executing the above steps, the method disclosed in this embodiment, through precise pre-delay compensation of material flow data across devices and combined with the weight reduction and purification mechanism of the dynamic disturbance model, fundamentally suppresses the over-compensation and violent oscillation phenomenon caused by torque misjudgment in the system.

[0180] Furthermore, by combining a smooth transition strategy based on disturbance classification with a decision-making logic based on energy consumption perception, this method significantly improves the displacement tracking accuracy, system robustness, and absolute stability of the paving vessel in complex and harsh sea conditions, while completely avoiding high-frequency wear of the actuators and maintaining a dynamic balance of energy consumption throughout the entire vessel's operation. This meets the stringent requirements of modern high-end underwater engineering equipment for intelligent micro-motion control.

[0181] Example 5:

[0182] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0183] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0184] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0185] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0186] The memory 103 stores a computer program corresponding to the adaptive shift control method for pavers based on flow feedforward according to the above embodiments of the present invention. This computer program is executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0187] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0188] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An adaptive displacement control system for a paving vessel based on flow feedforward, characterized in that, include: The collaborative data acquisition module is used to collect local data of the paving vessel and cross-device collaborative data in response to the paving vessel's movement command, and to perform delay compensation on the material flow data in the cross-device collaborative data. The dynamic disturbance model construction module is used to construct a dynamic disturbance model based on the local data of the paving vessel and the cross-equipment collaborative data after delay compensation, and to calculate and generate feedforward compensation amounts for the thrusters and anchor winches through the dynamic disturbance model. A two-stage adaptive decision-making module is used to generate control commands based on the feedforward compensation amount and feedback decoupling optimization logic to adjust the actual output of the side thruster and the anchor winch. The calculation process for delay compensation of material flow data in the collaborative data acquisition module includes: obtaining the preset material pipeline length, real-time material flow rate and sensor inherent delay, and calculating the total measurement delay; calculating the predicted material flow rate at the actual moment of the disturbance based on the current material flow rate change rate and the total measurement delay, and inputting the predicted material flow rate into the dynamic disturbance model.

2. The system according to claim 1, characterized in that, The process of calculating the total measurement delay in the collaborative data acquisition module includes: The fluid transmission delay is obtained by dividing the length of the material pipeline by the real-time material flow rate, and the fluid transmission delay is added to the inherent delay of the sensor to obtain the total measurement delay; The process of calculating the predicted material flow rate at the actual moment of the disturbance includes: multiplying the current material flow rate change rate by the total measurement delay to obtain the flow change compensation amount, and adding the real-time material flow rate at the current moment to the flow change compensation amount to obtain the predicted material flow rate.

3. The system according to claim 1, characterized in that, The process of constructing the dynamic disturbance model by the dynamic disturbance model construction module includes: Extract feature variables from the local data and cross-equipment collaborative data of the paving vessel. The feature variables include at least material flow rate, material pump outlet pressure, material concentration, nozzle attitude, draft, ocean current velocity, and wave height. Configure corresponding feature weight coefficients for each of the aforementioned feature variables; Each of the aforementioned feature variables is multiplied by its corresponding feature weight coefficient to obtain each feature component, and all the aforementioned feature components are weighted and summed to calculate the disturbance moment used to characterize the degree of disturbance to the paving vessel.

4. The system according to claim 1, characterized in that, The dynamic perturbation model construction module also includes a hybrid identification unit, used for: Clustering algorithms were used to classify the disturbances experienced by the paving vessel into flow-dominated disturbances, sea state-dominated disturbances, and coupled disturbances. For different types of disturbances, the corresponding parameter update algorithm is triggered to iteratively update the feature weight coefficients in the dynamic disturbance model.

5. The system according to claim 1, characterized in that, The process by which the two-stage adaptive decision module generates control commands includes: The feedforward compensation is decomposed into feedforward thrust and feedforward pull; Determine whether the feedforward thrust is greater than a preset thrust threshold and whether the feedforward pull is greater than a preset pull threshold; If the feedforward thrust is greater than the preset thrust threshold or the feedforward pull is greater than the preset pull threshold, then the value of the feedforward compensation amount is reduced by the preset attenuation step size, and a flow limit command is sent to the material pump. If the feedforward thrust is not greater than a preset thrust threshold and the feedforward pull is not greater than a preset pull threshold, then based on the currently allocated feedforward weight and feedback weight, the feedforward thrust and the feedback optimized thrust are weighted and fused to generate the final thrust command, and the feedforward pull and the feedback optimized pull are weighted and fused to generate the final pull command.

6. The system according to claim 5, characterized in that, The allocation process of feedforward weights and feedback weights in the two-order adaptive decision-making module includes: Real-time energy consumption is calculated based on the actual output thrust of the side thrusters, the shifting speed of the paving vessel, the actual output pulling force of the anchor winch, and the raising and lowering speed of the anchor winch. Calculate the energy consumption deviation between the real-time energy consumption and the preset optimal energy consumption reference value; In response to the condition that the energy consumption deviation value is greater than a preset deviation threshold and the current pose error of the paving vessel is less than a preset error threshold, the feedforward weight is reduced proportionally and the feedback weight is increased to maintain the sum of the feedforward weight and the feedback weight constant.

7. The system according to claim 1, characterized in that, The system also includes a traffic interference processing module, which is used to perform pulse-level filtering on cross-device collaborative data before the two-stage adaptive decision module generates control commands; The execution process of the pulsation hierarchical filtering includes: dividing the pulsation intensity into a first pulsation level, a second pulsation level, and a third pulsation level according to the variance of the material flow rate, and calling a moving average filtering algorithm with a different window width for different pulsation levels. Simultaneously, the filtering coefficients of the moving average filtering algorithm are dynamically updated based on the differential pressure value of the material pipeline.

8. The system according to claim 3, characterized in that, The dynamic perturbation model construction module also includes an extreme coupling decoupling unit, used for: The wave height change rate and the pressure change rate of the material pump outlet pressure are monitored in real time. In response to the operating conditions where the wave change rate is greater than a preset wave mutation threshold and the pressure change rate is greater than a preset pressure mutation threshold, an overcompensation suppression strategy is triggered. In the overcompensation suppression strategy, a first data sequence of wave height and a second data sequence of material pump outlet pressure within a preset historical time window are extracted, and the Pearson correlation coefficient between the two data sequences is calculated. The attenuation ratio of the characteristic weight coefficient for the outlet pressure of the material pump is calculated based on the Pearson correlation coefficient, wherein the value of the Pearson correlation coefficient is positively correlated with the magnitude of the attenuation ratio; The characteristic weight coefficient of the material pump outlet pressure is updated by reducing the weight using the attenuation ratio, and the disturbance torque is recalculated using the updated characteristic weight coefficient.

9. The system according to claim 4, characterized in that, The hybrid identification unit is configured with smooth transition update logic, and the execution process of the smooth transition update logic includes: When performing classification calculations using clustering algorithms, the distance values ​​from the current feature variable to the centers of each preset classification cluster are calculated, and the first distance value with the smallest value and the second distance value with the second smallest value are extracted. Calculate the ratio of the first distance value to the second distance value, and in response to the ratio being greater than a preset boundary confusion threshold, determine that the current system is in a disturbance type transition state; During the disturbance type transition state, the first parameter update algorithm associated with the first distance value and the second parameter update algorithm associated with the second distance value are started simultaneously to calculate the first model coefficient update amount and the second model coefficient update amount, respectively. The reciprocal of the first distance value is extracted as the first fusion weight, and the reciprocal of the second distance value is extracted as the second fusion weight. The first fusion weight and the second fusion weight are used to perform weighted fusion on the first model coefficient update amount and the second model coefficient update amount to generate the transition period update amount. The preset feature weight coefficients in the dynamic perturbation model are iteratively updated using the transition period update amount.

10. An adaptive shift control method for pavers based on flow feedforward, characterized in that, Includes the following steps: In response to the paving vessel's relocation command, local data of the paving vessel and cross-equipment collaborative data are collected, and delay compensation is performed on the material flow data in the cross-equipment collaborative data. Based on the local data of the paving vessel and the cross-equipment collaborative data after delay compensation, a dynamic disturbance model is constructed, and feedforward compensation amounts for the side thrusters and anchor winches are calculated and generated through the dynamic disturbance model. Based on the feedforward compensation and feedback decoupling optimization logic, control commands are generated to adjust the actual output of the side thruster and the anchor winch, and to execute the adaptive displacement of the paving vessel. The calculation process for delay compensation of material flow data in the cross-device collaborative data includes: obtaining the preset material pipeline length, real-time material flow rate and sensor inherent delay, and calculating the total measurement delay; calculating the predicted material flow at the actual time of the disturbance based on the current material flow rate change rate and the total measurement delay, and inputting the predicted material flow into the dynamic disturbance model.

Images